Carotenoid consumption is of great interest in disease prevention studies. Until recently, carotenoid food composition data have not been available from a single laboratory source with high validity/reliability characteristics. With the availability of a new carotenoid food composition data base, we examined the impact of the new data base on the intake estimates as measured by a food frequency questionnaire and on the relationship of those estimates to plasma values to ascertain what, if any, improvement is achieved through use of the new values. Plasma samples were available for 162 healthy adults participating in cancer prevention studies at the Arizona Cancer Center, including men and women, smokers and nonsmokers. A single laboratory analyzed plasma samples for beta-carotene, alpha-carotene, lutein, and lycopene. All subjects had completed a modified version of the Block food frequency questionnaire, which calculates carotenoids using a literature-based algorithm. A new carotenoid composition data base using recently published data (A.R. Mangels et al., J. Am. Diet. Assoc., 93: 284-296, 1993) was then directly substituted for the Block data base. There were high correlations between intake estimates derived from the two data bases for all four carotenoids (range, r = 0.76-0.96). Average intake estimates based on the Mangels et al. data base were significantly higher for beta-carotene and lycopene; however, correlations between intakes and plasma values were significantly different only for beta-carotene (r = 0.44 for Mangels versus 0.32 for Block, P = 0.015).